Claude Code v2.1.282 ·
Claude Code now records thinking and temperature support for each model
Claude Code's per-model capability details now include whether a model supports extended thinking and whether it accepts a temperature
Under the hoodTier: how much it should matter to you
2Useful: my rating, 1 to 5
3Signal: worth watching, 1 to 5
ModelsArea: what it touches
Internal ChangesKind: in v2.1.282,
Internal ChangesSection of the release
What
Claude Code keeps a record of what each model can do. That record now also notes:
- whether the model supports thinking (reasoning before it answers)
- whether it refuses requests that turn thinking off
- whether it limits effort when thinking is turned off
- whether it accepts a temperature setting, which controls how varied its answers are
Why
In practice, this information may appear in model details or debugging output.
Something disagreesSomething we can check disagrees with this entry, or the writer said they could not settle it.
The writer flagged doubtIt is not clear what uses this record or whether any user sees it.
See this entry in the whole of v2.1.282 →
